Class A/B net current when push and pull are on

[crunchynut]

Hi all,

Apologies if this is a stupid question and/or the wrong place to ask it, but I guess answers will be easily provided by you experts, so here goes.

I broadly get how a class a/b amp works (in very basic terms), but something is puzzling me.

During the part of the cycle where both output transistors are on, so one is pushing current into the load and one is pulling current from the load, then why isn’t the net current in the load during this period zero? I know it can’t be else it wouldn’t work, but why isn’t it? I’m guessing the answer is that my understanding is way too basic, but please try and educate me …

Thanks

[James Perrett]

We're not talking about a digital system. The output transistors will gradually switch on or off depending upon their input current and they will only switch fully on if you are clipping. If they are exactly balanced then the output will be 0. There is normally a quiescent current of a few mA through the output transistors when the amp is at rest - this current is often adjustable to match the transistors fitted.

[Wonks]

....or valves.

[Hugh Robjohns]

During the part of the cycle where both output transistors are on, so one is pushing current into the load and one is pulling current from the load, then why isn’t the net current in the load during this period zero?

The current to the load is zero...

When the input signal is 0V, both transistors are passing a small current, but the positive rail transistor's current is exactly the same as the negative rail transistors current. In effect, one transistor is feeding the other and there's nothing left to feed the load. So zero input equally zero output.

As the input signal increases, one transistor starts to turn on more and thus passes excess current into the load, while the other turns off more taking less current from the other transistor, leaving more for the load. This is essentially class-A operation.

With larger signals one transistor will be completely turned off, and the other turned on more, to operate like the Class-B amp.

[crunchynut]

Ah, thank you, that makes sense with what I was thinking about transistor currents offsetting each other.

But, with a rising input signal (say), during that small period before the pull transistor turns off and during which current being sourced by the push transistor is being sunk by the pull transistor (but by a decreasing amount until it’s zero), do you have non linearity in the output and hence distortion?

[Hugh Robjohns]

No, the whole point is that there is no distortion because the biasing ensures both transistors are operating in Class-A over that small region.

[merlyn]

You may be confusing yourself with the idea of push and pull. In a class B amp one transistor does the positive swing, and one transistor does the negative swing. That's it.

This graphic shows how amps are classified :

[crunchynut]

No, the whole point is that there is no distortion because the biasing ensures both transistors are operating in Class-A over that small region.

But that’s what confuses me. If both transistors are in class A in that small region, so both are conducting, then doesn’t the pull transistor sink the push transistor’s current, so there is no load current?

[Hugh Robjohns]

The 'pushing and pulling you're thinking of is strictly Class-B operation.

For very small signals in a class-AB setup, where both transistors are biased into conduction, they essentially work in Class-A, and effectively in parallel — such that they both push and pull (some of the) current into/from the load.

In reality, one will be doing slightly more work than the other depending on whether the signal is going positive or negative, but they are both contributing.

There's lots of explainers of electronics like this on the web or in books if you're keen to learn more about it. I can recommend Douglas Self's book on small signal Audio design as a really good primer for audio electronics in general.

https://www.soundonsound.com/reviews/sm ... nd-edition

He has also written a book specifically on power amp design called, Audio Power Amplifier Design...

[Jimmy B]

No, the whole point is that there is no distortion because the biasing ensures both transistors are operating in Class-A over that small region.

But that’s what confuses me. If both transistors are in class A in that small region, so both are conducting, then doesn’t the pull transistor sink the push transistor’s current, so there is no load current?

Here's another way to look at it:
If both transistors are conducting equally, then yes, "the pull transistor exactly sinks the push transistor's current", and the net output is zero.
If, however, one transistor is conducting more than the other, then the difference current flows in the load. This is the output current of the amplifier in that limited class A region of operation. The rest of the amplifier circuit is what makes sure that all of this works in an accurate way.
A very simplified view, but it may help.

[merlyn]

In the interests of preserving the meaning of engineering terms, "class A in a small region" doesn't make any sense.

[Wonks]

My explanation (from reading a few articles on the web, so I may not have it quite right)

Each half of the circuit in a class AB amp is capable of providing both a positive and negative voltage/current signal. So it's not one side is always 'push' and the other 'pull'. Both sides can push and pull. The amount to which the 'push' side can 'pull' and the 'pull' side can 'push' depends on the amount of DC bias applied (at least in a transistor amp).

Given enough bias, each side could be fully capable of providing the full waveform, but then each side would be acting as a class A amp.

So each side should have just enough bias added to allow it to do a bit of the other half of the cycle to overcome (on transistors) the roughly 0.7v it requires to get the transistor to start conducting.

During the period when both halves are conducting, each half is putting out the same voltage, so if 'push' is outputting 0.5v, then pull is also outputting 0.5v.

If 'pull' is outputting -0.5v, then 'push' is also outputting -0.5v. So any current flowing should flow equally from the two halves into the load (speaker) or from the load into the two halves of the amp circuit.

There may be a slight imbalance in voltage and current flow if the two halves aren't biased exactly the same way or one transistor has a slightly higher impedance than the other, but in a perfect amp, you don't get current flowing from 'push' to 'pull', or vice versa. And even with small imbalances, most of the current is flowing through the load.

Then once the input signal to the 'push' or 'pull' circuit rises (or falls) below the bias threshold for transistor operation, then the output from that circuit stops, and the other circuit works on its own.

You can alter the bias voltage so that each half conducts for a lot more than a nominal 181°-200° of the cycle, but this moves it closer to full class A operation in terms of efficiency.

But you might want to do this so that with small signal levels, the amplifier always operates with both sides conducting, and so operates in pure class A mode. When the input signal gets bigger and the bias voltage level is exceeded, by the nature of how it works, it switches over to class AB operation.

(Goes and hides behind the sofa).

[merlyn]

... During the part of the cycle where both output transistors are on, so one is pushing current into the load and one is pulling current from the load, ...

I think the above is where the confusion is arising. The load (a speaker) doesn't produce current. It's always the amp that produces current, but because it's AC it goes in two directions, which is your 'push' and 'pull'.

More technically a power amp can be thought on as a voltage controlled current source, as contrasted with a preamp, which can be thought on as a voltage controlled voltage source.

[Nazard]

He has also written a book specifically on power amp design called, Audio Power Amplifier Design...

My copy, 6th Edition, arrived yesterday - a great resource, even though I have all the EWW articles he wrote. Plus some very well made and amusing points about 'subjectivism'. But since I am planning to build one of his power amps to go with my 2012 Elektor design, it made good sense to have more information at hand.

[ef37a]

Crunchynut, are you familiar with the concept of balanced outputs and inputs?
Hope so because the output stage of a transistor amp is a 'balanced' system.

There are two equal supplies but of opposite polarity and two (or more in ll) transistors. The DC negative feedback forces the mid point of the transistors to be zero volts within a few mV. Since the load is connected between mid point and supply zero, no current can flow.

If the upper transistor's base (assuming an NPN device) is made more positive current will flow from that, through the load to zero V. When the bottom PNP transistor's base is made negative current flows through that device, through the load to 0V.

The "problem" comes at the point where one transistor is near to zero current and is 'handing over' to the other one, also passing a low current. This is "crossover Distortion" and is THE main source of distortion in amplifiers other that pure class A. That is horribly inefficient and in an age when a 100W amp is considered a 'norm' simply not practical.

Douglas Self's amp book is excellent and I commend it to all (although I confess quite a lot of it was beyond me!) He has no time for "class AB" amps as he has shown that they actually produce MORE distortion over the whole power range than what he defines as "class B". That is an amplifier where Xover products reach a minimum at some bias point. Biasing 'hotter' than that produces no overall improvement. MOSFETs also get short shrift due to their low gm.

He describes designs that give astonishingly low distortion, in the third decimal place up to 20kHz and some class B designs are so good that they have to be fitted with a 'jumper' to make the distortion measurable and allow bias to be set!

He is also not at all keen on the then class D designs but that was some ten years ago. He suggests that class G can get close to the efficiency of D but with vastly better linearity and is it interesting that at least one top monitor maker uses class G and not a switcher?

Dave.

[crunchynut]

Thank you all for taking the time to explain. Getting different explanations is really helpful in building a clear picture in my mind as to how they work. Brilliant. Thank you.

[Jimmy B]

I still listen to class AB amplifers, in monitors and room Hi Fi, because that is what I have, and because they are good enough that I see no need for sonic improvements. This is in part due to the efforts of Mr Self.
However, I am becoming increasingly aware that class AB amplifiers get warm, and are therefore costing me money in wasted energy. Is the time approaching when I should think about getting class D amplifiers instead? If so, does it make this discussion moot?

[Folderol]

I still listen to class AB amplifers, in monitors and room Hi Fi, because that is what I have, and because they are good enough that I see no need for sonic improvements. This is in part due to the efforts of Mr Self.
However, I am becoming increasingly aware that class AB amplifiers get warm, and are therefore costing me money in wasted energy. Is the time approaching when I should think about getting class D amplifiers instead? If so, does it make this discussion moot?

Not at all. There could easily be circumstances where the HF switching components could cause trouble.

[Jimmy B]

Yes, I heard that there were a lot of complaints about this and other aspects of the Sinclair X-10 in the 1960s. Hopefully things have moved on a bit since then.
I did try a Tripath when they were all the rage, and while it was remarkable for its size and cost, it couldn't compete with a real amplifier. It did drive my desktop speakers for a while until I found something better.

[ef37a]

I still listen to class AB amplifers, in monitors and room Hi Fi, because that is what I have, and because they are good enough that I see no need for sonic improvements. This is in part due to the efforts of Mr Self.
However, I am becoming increasingly aware that class AB amplifiers get warm, and are therefore costing me money in wasted energy. Is the time approaching when I should think about getting class D amplifiers instead? If so, does it make this discussion moot?

There are two types of class B (Selfs) output stages. The emitter follower and the Complimentary Feedback Pair (CFP) The former needs a bias current of around 100 to 150mA which means a 50W per ch amp would still dissipate some 10W per channel. The CFP requires a tenth that current and so are much more efficient.
CFP OP stages are very marginally poorer performers for distortion in some areas but still in the third decimal place to 20kHz. I know some very low THD figures are claimed for class D amps but since there seems to be no truly independent magazine testing amps these days I would still bet on the Self designs!

Self also shows that there are techniques that can produce even lower distortion, including crossover distortion but the amplifiers cannot be relied upon to be stable for any gash load that come down the pike! However, contained in an active speaker they would be just about the last word in fidelity.

Dave.

[Wonks]

However, just about all active monitors use class D amps these days.

[ef37a]

However, just about all active monitors use class D amps these days.

Most but not all Wonks and AFAIK no one has every stuck one on an AP rig?

Dave.

[Jimmy B]

I think that I should apologise to crunchynut for hijacking his discussion. I have seen other people do it and I hadn't realised how easy it is to wander off topic.

Sorry.

[Wonks]

However, just about all active monitors use class D amps these days.

Most but not all Wonks and AFAIK no one has every stuck one on an AP rig?

Dave.

AP rig?

[Drew Stephenson]

However, just about all active monitors use class D amps these days.

Most but not all Wonks and AFAIK no one has every stuck one on an AP rig?

Dave.

AP rig?

Audio Precision test rig?

I think that I should apologise to crunchynut for hijacking his discussion. I have seen other people do it and I hadn't realised how easy it is to wander off topic.

Sorry.

I reckon it's ok for subjects to wander once the initial query has been answered, and I think Crunchnut had got what they were after.